A dynamic response user preference home product intelligent recommendation method

Abstract

The application discloses a kind of dynamic response user preference home product intelligent recommendation method.It includes the following steps: the home product style attribute is extracted as home style feature vector;User preference is obtained and characterized as the input of recommendation algorithm;Adopt double strategy combination mode, dynamically respond to the preference of user, set strategy start condition, calculate the degree of user between the coordination, the satisfaction of user to home product and the coordination degree between different categories of home product, and intelligently recommend home product for user.The application obtains user preference dynamically and analyzes the relationship between it and home product attribute, considers the style coordination problem between home products, intelligently recommends suitable home product for user, reduces the time and effort required for user to select home product independently.

Description

Technical Field

[0001] This invention belongs to the field of personalized home furnishing recommendations, specifically relating to a method for intelligently recommending home furnishing products that dynamically responds to user preferences. Background Technology

[0002] Home furnishing products have become a highly sought-after sector in recent years. As people's living standards continue to improve, their demand for comfortable, warm, and aesthetically pleasing home environments is also increasing, with a desire for home furnishing products that better suit their personal lifestyles and style preferences. However, with continuous technological advancements, the variety of home furnishing products has also proliferated, often making selection overwhelming and difficult. Therefore, in the field of home furnishing recommendations, how to recommend products based on user preferences has become a pressing issue that needs to be addressed.

[0003] Traditional home furnishing product recommendation methods have several problems. First, these methods often only consider users' historical behavior, neglecting their preferences and specific needs in different scenarios. Second, they often focus on only one brand or product category, ignoring the fact that users may have multiple choices. Furthermore, users often need to spend a significant amount of time and effort selecting suitable products from product images and descriptions, causing considerable inconvenience. These issues result in inaccurate and inefficient recommendations, failing to meet users' personalized needs.

[0004] Dynamic response technology refers to the timely adjustment and improvement of products or services based on users' real-time needs and feedback to meet their personalized requirements. In the field of recommendation systems, dynamic response technology can make recommendations more targeted and practical, thereby improving user satisfaction and the shopping experience.

[0005] This method addresses the problems of traditional methods by proposing a dynamic, user-responsive intelligent recommendation method for home furnishing products, aiming to recommend home furnishing products to users more intelligently and efficiently. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of existing technologies by providing a method for intelligently recommending home furnishing products that dynamically responds to user preferences. This method employs a dual-strategy combination mode to dynamically respond to user preferences. By calculating the compatibility between users and the recommendation score of the home furnishing products to be recommended, it recommends home furnishing products to users, thereby achieving intelligent and efficient recommendation of home furnishing products.

[0007] To achieve the above objectives, the present invention provides a smart home product recommendation method that dynamically responds to user preferences, comprising the following steps:

[0008] (1) Extract the style attributes of home furnishing products into a home furnishing style feature vector S I ;

[0009] (2) Obtain user preferences and extract them as user expectation vector U V U V Including the current room type R and the user's style preference S U And the category preference for home furnishing products, Q;

[0010] (3) For each home furnishing category in category preference Q, the dual-strategy combination mode dynamically responds to user preferences and recommends home furnishing products of the current category to the user. The user selects a model from the recommendation list and adds it to the recommendation space.

[0011] (4) Repeat step (3) until the user completes the selection of all categories in category preference Q, then the recommendation is completed and all furniture models in the recommendation space are output.

[0012] In step (1):

[0013] The style attributes of home furnishing products are extracted into a home furnishing style feature vector S. I The dimension is 1×M;

[0014] The style feature vector of product with model number i is s im Let m represent the probability that product i is biased towards style m, where m∈M and M is the total number of style types.

[0015] In step (2): the user expectation vector U V Used to store user preferences, including the current room type R and the user's style preference S. U And the category preference for home furnishing products, Q;

[0016] (a) Label and encode different types of rooms, where R is the label and encoding result of the current room type;

[0017] (b) User style preferences S U For the user's style preference for the current room, and the home style feature vector S I The dimensions and meanings are the same; S U ={s u1 ,s u2 ,…,s um}, s um This represents the user's preference coefficient for style m;

[0018] (c) The category preference Q has a dimension of 1×N, where N is the total number of home furnishing product categories, and Q={q u1 ,q u2 ,…,q un}, q unThis represents the user's preference score for home furnishing category n, where n∈N. Home furnishing categories include sofas, beds, televisions, lamps, dining tables, etc.

[0019] The specific steps (3) are as follows: the current room R is abstracted as a recommendation space Θ. Each dynamic response will recommend multiple models of home furnishing products to the user, and the user can select one of them to put into the recommendation space Θ.

[0020] The system sequentially queries the home furnishing categories corresponding to the non-zero elements in the user's category preference Q, and uses these categories as the recommended home furnishing categories for each dynamic response. Each dynamic response employs a dual-strategy combination to present the recommended list RL to the user. i The recommendation list contains multiple models in various home furnishing categories to be recommended. Users can select from the recommendation list RL i Choose a favorite model and add it to the recommended space Θ;

[0021] The dual-strategy combination mode includes strategy A and strategy B. Each dynamic response combines the recommendation results of strategy A and strategy B to output a home furnishing recommendation list Rl. i .

[0022] Strategy A is as follows:

[0023] 3.1) Use the same room type R as the current user as the filtering condition to perform preliminary filtering of historical users (find historical users with the same room type as the current user) and obtain the filtered users B that match the current user.

[0024] 3.2) When the number of users to be filtered is greater than max{M,N}, strategy A is activated to recommend home furnishings to the users, i.e., step 3.3) is executed; otherwise, strategy A is not activated and strategy A does not output recommended home furnishing product models.

[0025] 3.3) Calculate the compatibility between current user A and each filtered user B. The calculation formula is as follows:

[0026]

[0027] Where, V∈(S U ,Q);

[0028] When V = S U hour, Distance based on home style preferences;

[0029] Among them, S U (A) is the style preference vector of user A. It is the weight of style preference distance;

[0030] When V = Q Category preference distance for home furnishing products;

[0031] Where Q(A) is user A's category preference vector for home furnishing products, w Q It is the weight of the category preference distance.

[0032] 3.4) Based on the calculation results, add the home furnishing product models selected by historical users with a compatibility level ≥ the activation threshold C to the home furnishing recommendation list RL. i If there are more than 3 historical users whose compatibility is greater than or equal to the activation threshold C, only the home furnishing product models selected by the top 3 historical users with the highest compatibility will be used.

[0033] If there are no historical users whose compatibility is greater than or equal to the activation threshold C, strategy A will not recommend home furnishing products.

[0034] Strategy B is as follows:

[0035] Calculate the recommendation score for all models in the current home furnishing category to be recommended, including the current user's satisfaction with the recommended home furnishing products, and the coordination between the home furnishing products to be recommended and the selected home furnishing products in the recommended space Θ. The calculation formula is as follows:

[0036]

[0037] in, The current user's satisfaction with the recommended home furnishing product i;

[0038] The degree of coordination between the home furnishing product to be recommended (i) and all selected home furnishing products in the recommended space (Θ);

[0039] Where i represents a product model in the home furnishings to be recommended, and m represents a product model in the recommended space Θ; and q are the home style feature vectors for product i and product m, respectively; ui and q um These are the user's preference scores for the home furnishing categories of product i and product m, respectively; w s The weighting of current users' satisfaction with different home furnishing models is calculated using the following formula:

[0040]

[0041] In the formula, n is the total number of home furnishing product types in the recommendation space Θ;

[0042] Based on the calculation results, the top 3 home furnishing product models with the highest recommendation scores in the home furnishing category will be added to the home furnishing recommendation list RL. i middle.

[0043] Step (4) is to repeat step (3) until all non-zero elements in the user category preference Q are traversed, and then the recommendation is completed.

[0044] The beneficial effects of this invention are as follows:

[0045] This invention obtains users' style and category preferences for home furnishing products, extracts the style features of home furnishing products, adopts a dual-strategy combination mode, dynamically responds to user preferences, and recommends home furnishing products to users by calculating the compatibility between users and the recommendation score of the home furnishing products to be recommended.

[0046] The method of this invention is user-centric and user-selection-oriented, dynamically responding to user satisfaction and the harmony of home furnishing product styles. This improves the efficiency and accuracy of recommendations, allowing users to better experience the intelligent recommendation service for home furnishing products and reducing the time and effort required for users to independently select home furnishing products. Attached Figure Description

[0047] Figure 1 This is a flowchart of the intelligent home product recommendation method that dynamically responds to user preferences according to the present invention. Detailed Implementation

[0048] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the invention as detailed in the appended claims.

[0049] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The singular forms “a,” “the,” and “the” used in this invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0050] It should be understood that although the terms first, second, third, etc., may be used in this invention to describe various information, this information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first information may also be referred to as second information without departing from the scope of this invention, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0051] The specific embodiments of the present invention will now be described more completely and clearly with reference to the accompanying drawings. Figure 1A method for intelligently recommending home products in response to user preferences includes the following steps:

[0052] (1) Extract the style attributes of home furnishing products into a home furnishing style feature vector S I The dimension is 1×M, where M is the total number of style types; specifically: assuming there are M home furnishing styles, denoted by s im Let represent the preference coefficient of the current product model i for style m. Then, the style feature vector of this model is represented as: s im The range is between [0,1], s im The larger the value, the greater the proportion of style m in the current product model i. Therefore, the home furnishing style feature vector of this model can be S. I ={0,0.75,0,0,0,0.25}.

[0053] (2) Obtain user preferences and extract them as user expectation vector U V ={R,S U ,Q},U V This includes the current room type R and the user's style preference S for each room. U And the category preference for home furnishing products, Q; specifically:

[0054] The current room type R uses a label encoding method. For example, if the room type includes living room, bedroom, kitchen, bathroom, and balcony, the label encoding results would be: living room - 0, bedroom - 1, kitchen - 2, bathroom - 3, balcony - 4. Assume there are M different home decor styles, denoted by s. um Let S represent the user's preference coefficient for style m. Then, the user's style preference is represented as S. U ={s u1 ,s u2 ,…,s um}, s um The range is between [0,1], s um The larger the value m, the more the user prefers that style. Assuming there are n home furnishing categories, let q represent this preference. un If we represent the user's category preference score, then the category preference for home furnishing products is represented as Q = {q}. u1 ,q u2 ,…,q un}, q un The range of q is between [0,1]. un The larger the value, the more the user prefers home furnishing category n. Therefore, the user expectation vector can be U. V ={{1},{0,0,0.8,0,0.2,0},{0,0.5,0.8,1,0.2,0,0,0.1,0.6,0}}.

[0055] (3) A dual-strategy combination model is adopted to dynamically respond to user preferences and recommend home furnishing products to users. Strategy A calculates the compatibility between the current user and other users; Strategy B calculates the recommendation score of home furnishing products, including the current user's satisfaction with different home furnishing product models and the coordination between the selected and unselected home furnishing products; specifically:

[0056] The current room R is abstracted as a recommendation space R. Each dynamic response recommends multiple furniture product models to the user, who then selects one to add to the recommendation space R. The furniture categories corresponding to the non-zero elements in the user's category preference Q are sequentially queried and used as the furniture category to be recommended in a single dynamic response. To avoid cold start, a dual-strategy combination mode is used to recommend furniture product models to the user. The dual-strategy combination mode is as follows:

[0057] Strategy A: Find compatible users for the current user and use the same current room type R as a filtering condition to perform preliminary filtering of historical users in order to reduce the amount of computation.

[0058] For example, the current user's expectation vector U V (A) = {{1}, {0, 0, 0.8, 0, 0.2, 0}, {0, 1, 1, 3, 2, 0, 0, 1, 1, 0}}, where U is the user expectation vector of a certain historical user. V If (B) = {{1}, {0,0,0.7,0,0.3,0}, {0,1,0,2,2,0,1,0,1,1}}, then user B can be used as a filter user.

[0059] After completing the initial screening, calculate the compatibility between current user A and screened user B using the following formula:

[0060]

[0061] Where V∈(S) U Q). When V = S U At that time, calculate the style preference distance. This is the weight of the style preference distance; when V = Q, the home furnishing preference distance is calculated, w Q It is the weight of the distance to home preference.

[0062] To ensure the covariance matrix in the formula is invertible, the number of users to be screened should be greater than S. U The dimension of Q is max{M,N}. If the number of historical users is small and this condition cannot be met, strategy A for recommending home furnishings to users will not be activated; however, when the number of historical users is large, the number of users after filtering can easily meet this condition.

[0063] Based on the calculation results, if there are historical users whose compatibility is greater than or equal to the activation threshold C, then the furniture models selected by the top 3 historical users with the highest compatibility will be added to the furniture recommendation list RL. i If yes, then strategy A will not be activated to recommend home furnishing products to the user; otherwise, strategy A will not be activated.

[0064] For example, assuming the activation threshold C = 0.65, S in the above... U The dimension of Q is 6, and the dimension of Q is 10. Therefore, when the number of users to be screened is greater than 10, and there are screened users with a compatibility score of ≥0.65, then strategy A will be activated to recommend home furnishings to the users.

[0065] Strategy B: Calculate the recommendation score for all models in the furniture category to be recommended, including the current user's satisfaction with different furniture models, and the coordination between the furniture to be recommended and the selected furniture products in the recommended space Θ. The calculation formula is as follows:

[0066]

[0067] Where i represents a product model in the home furnishings to be recommended, and m represents a product model in the recommended space Θ; q ui and q um These are the user's preference scores for the home furnishing categories i and m, respectively; w s To calculate the weight of current users' satisfaction with different home furnishing models, the calculation formula is as follows:

[0068]

[0069] Where n is the total number of home furnishing product types in the recommendation space Θ. As n increases, the weight of the current user's satisfaction with home furnishing model i decreases, while the weight of the coordination between the home furnishing to be recommended and the selected home furnishing products in the recommendation space Θ increases. For example, when n = 0, it means that the first type of home furnishing product is being recommended to the user, w s =1, meaning only satisfaction weight is considered; when n=3, it means that the user has already selected 3 home furnishing categories in the current recommendation space Θ, and a 4th home furnishing category needs to be recommended to the user.

[0070] Based on the calculation results, the top 3 recommended furniture models will be added to the furniture recommendation list RL. i middle.

[0071] Finally, combining the recommendation results of Strategy A and Strategy B, the home furnishing recommendation list RL for this dynamic response is generated. i The system displays a list of models that the user can select and add to the recommended space.

[0072] (4) Repeat step (3) until all non-zero elements in user category preference Q are traversed, then the recommendation is complete.

[0073] In each dynamic response, the furniture selected by the user will be placed in the recommendation space Θ. Subsequent recommended furniture will be combined with the selected furniture products in the recommendation space Θ and the user's style preference S. U Calculate recommendation scores to dynamically respond to user preferences.

Claims

1. A method for intelligently recommending home furnishing products in response to user preferences, characterized in that, Includes the following steps: (1) Extract the style attributes of home furnishing products into home furnishing style feature vectors. ; (2) Obtain user preferences and extract them as user expectation vectors , Including the current room type User style preferences and preferences for home furnishing product categories ; (3) Category preference Each home furnishing category in the system dynamically responds to user preferences through a dual-strategy combination mode, recommending home furnishing products of the current category to users. Users can then select a model from the recommendation list to add to their recommendation space. (4) Repeat step (3) until the user completes the category preference. If all categories are selected, the recommendation is completed, and all furniture models in the recommendation space are output; Step (3) specifically involves: Sequential query of user category preferences The home furnishing category corresponding to the zero element in the non-Chinese context is used as the home furnishing category to be recommended in a single dynamic response. Each dynamic response uses a dual-strategy combination mode to present a recommendation list to the user. The recommendation list contains multiple models in various home furnishing categories to be recommended. Users can select from the recommendation list... Choose a favorite model and add it to your recommended space. middle; The dual-strategy combination mode includes Strategy I and Strategy II. Each dynamic response combines the recommendation results of Strategy I and Strategy II to output a home furnishing recommendation list. ; Strategy I is as follows: 3.1) Change the current room type Using similarity as a filter criterion, a preliminary screening of historical users is performed to obtain users who are compatible with the current user. ; 3.2) When the number of users to be filtered is greater than When the user is not in a certain situation, Strategy I is activated to recommend home furnishings (i.e., step 3.3 is executed); otherwise, Strategy I is not activated and does not output recommended home furnishing product models. in, The total number of style types. This represents the total number of home furnishing product categories. 3.3) Calculate the current user and each filtered user The compatibility between them is calculated using the following formula: in, ; when hour, Distance based on home style preferences; in, Let A be the style preference vector for user A. Let B be the style preference vector. It is the weight of style preference distance; when hour, Category preference distance for home furnishing products; in, Let A be the category preference vector for home furnishing products. Let's define the category preference vector for user B's home furnishing products. It is the weight of the category preference distance; 3.4) Based on the calculation results, the compatibility will be... Startup threshold Add the home furnishing product models selected by historical users to the home furnishing recommendation list. In the middle, if the compatibility is good Startup threshold For users with more than 3 historical users, only the home furnishing product models selected by the top 3 historical users with the highest compatibility are considered; If there is no compatibility Startup threshold For historical users, Strategy I will not recommend home furnishing products; Strategy II is as follows: Calculate the recommendation score for all models in the current home furnishing category to be recommended, including current user satisfaction with the recommended home furnishing products, as well as the home furnishing products to be recommended and the recommended spaces. The formula for calculating the coordination between the selected home furnishing products is as follows: in, Recommended home furnishing products for current users Satisfaction level; Home Furnishing Products to be Recommended and Recommendation Space Coordination among all selected home furnishing products; in, This is one of the product models in the home furnishings to be recommended. For recommendation space One of the product models; and Each product and products Home style feature vector; and users' opinions on the product and products Preference score for the relevant home furnishing category; The weighting of current users' satisfaction with different home furnishing models is calculated using the following formula: In the formula, For recommendation space The total number of home furnishing product categories in China; Based on the calculation results, the top 3 home furnishing product models with the highest recommendation scores in the home furnishing category will be added to the home furnishing recommendation list. middle.

2. The intelligent recommendation method for home furnishing products based on dynamic response to user preferences according to claim 1, characterized in that: In step (1): Extracting the style attributes of home furnishing products into home furnishing style feature vectors , dimension ; Model number The style feature vector of the product is , , Indicates model number Product style The probability, , This represents the total number of style types.

3. The intelligent recommendation method for home furnishing products based on dynamic response to user preferences according to claim 1, characterized in that: In step (2): User expectation vector Used to store user preferences, including the current room type. User style preferences and preferences for home furnishing product categories ; (a) Label and code different types of rooms. The label encoding result for the current room type; (b) User style preferences For users' style preferences for the current room, and home style feature vectors The dimensions and meanings are the same; , Indicates user's preference for style Preference coefficient; (c) Category preference The dimension is , This represents the total number of home furnishing product categories. , Indicates user preference for home furnishing categories Preference scores .

4. The intelligent recommendation method for home furnishing products based on dynamic response to user preferences according to claim 1, characterized in that: Step (4) is to repeat step (3) until user category preferences are traversed. If all non-zero elements are found, the recommendation is complete.

5. A computer storage medium, characterized in that, The computer storage medium includes a stored program, wherein, when the program is executed, it controls the device where the computer storage medium is located to execute the intelligent home product recommendation method that dynamically responds to user preferences as described in any one of claims 1 to 4.

6. A processor, characterized in that, The processor is used to run a program, wherein the program executes the intelligent home product recommendation method that dynamically responds to user preferences as described in any one of claims 1 to 4.

Images